Neural Network Based Estimation and Modelling of a Fusion Bioreactor

A bioreactor is a specialised vessel that allows for the growing of cells under sterile conditions while still maintaining control over the surrounding environmental conditions. Bioprocesses are used in the production of a wide range of goods in the food, beverage, and pharmaceutical sectors, among other things. Bioprocesses have complex dynamics, making their management a tough and sensitive operation; they are also intrinsically concerned with high nonlinearity, which makes modelling and parameter estimation extremely difficult tasks in this context. Furthermore, because to a paucity of online measurements of biomass and substrate concentrations, this job is more difficult to do. In order to properly regulate the process, an efficient model must be used to do this. An improved performance may be obtained by combining elements of more than one kind of model in a hybrid form. Combining the first principles model from the literature with the data acquired from the real-time bioreactor setup results in a hybrid model that may be used in a variety of applications. With the help of neural networks, the parameters of the process may be calculated more accurately. The newly constructed hybrid model is capable of forecasting the biomass concentration value, and this value is shown to be correlated with the real-time response of the bioreactor.